publish-content

name	publish-content
description	Publish approved content to social media platforms. Activate when user asks to publish, post, or distribute content. Trigger: "发布" / "发到小红书" / "帮我发" / "发布这篇". Supports Xiaohongshu, Douyin, WeChat Video, WeChat MP (公众号).

Publish Content

Utility skill. Dual-mode: clipboard-first (Free) + browser automation (Pro/OpenClaw with MCP Browser).

Tier Routing

Tier	Flow	How it works
Free	Clipboard-first	autocrew_publish action="clipboard" → format → user copies → opens platform → pastes → autocrew_publish action="confirm_published"
Pro	Browser CDP	OpenClaw MCP Browser automates the full publish flow end-to-end

Default: always use clipboard flow. Only use browser automation when MCP Browser is explicitly available and user opts in.

Clipboard Flow (Free — Default)

1. autocrew_publish action="clipboard" content_id="xxx" → Returns ClipboardOutput: formattedTitle, formattedBody, copyText, publishUrl, tips
2. Present the formatted content to user with copy instructions
3. User copies text, opens publishUrl, pastes, and publishes manually
4. autocrew_publish action="confirm_published" content_id="xxx" publish_url="" → Marks content as published with timestamp

Capabilities

Feature	OpenClaw	Claude Code
Pre-publish validation	✅	✅
Format for copy-paste	✅	✅
Browser automation publish	✅ (MCP Browser)	❌ (manual)
Status tracking	✅	✅

CRITICAL RULES (Browser Automation)

1. Navigate by URL, not by clicking menus (menus are unreliable in accessibility tree).
2. Wait 5 seconds after each navigation for page to load.
3. Video upload uses input[type=file] — find the hidden file input via evaluate, then use setInputFiles. NEVER try drag-and-drop.
4. Cover image upload — same approach: find input[type=file] for cover, use setInputFiles.
5. Truncate content to platform limits silently. Add "…" if truncated.
6. Respond in Simplified Chinese for all user-facing messages.

Platform Routes

Platform	Upload/Create URL	Login Detection
xiaohongshu	https://creator.xiaohongshu.com/publish/publish	URL contains /login
douyin	https://creator.douyin.com/creator-micro/content/upload	URL redirects to sso.douyin.com
wechat_video	https://channels.weixin.qq.com/platform/post/create	URL redirects to /login.html
wechat_mp	https://mp.weixin.qq.com/	URL contains /login or no cookie

Platform Limits

Platform	Title limit	Body limit	Tags
xiaohongshu	20 chars	1000 chars	5-15
douyin	30 chars	1000 chars	5
wechat_video	— (描述 only)	1000 chars	5
wechat_mp	64 chars	20000 chars	—

---

Step 0: Identify content to publish

1. List content: autocrew_content action="list"
2. Filter for status="approved" (or "draft" if user explicitly asks)
3. IF no content → tell user to write some first.
4. Load the content project: autocrew_asset action="list" content_id=xxx to check for cover/video assets.

Step 1: Pre-publish validation

Check content against platform limits. Auto-truncate with "…" if over limit.

Check required assets:

• xiaohongshu image note: needs at least 1 image
• douyin: needs video
• wechat_video: needs video
• wechat_mp: text only is fine

IF missing required assets → ask user to provide them, or use autocrew_asset action="add" to register.

Step 2: Route by mode

• IF browser automation available (OpenClaw MCP Browser) AND user wants auto-publish → go to Browser Automation Steps
• ELSE → go to Copy-Paste Mode

---

Copy-Paste Mode (Universal — uses clipboard tool)

Format content via autocrew_publish action="clipboard" content_id="xxx". The tool returns a ClipboardOutput object with platform-optimized text. Present it to the user like this:

Xiaohongshu format:

📋 小红书发布内容：

标题：{title}

正文：
{body}

标签：#tag1 #tag2 #tag3 #tag4 #tag5

📎 封面：{cover asset path or "需要手动上传"}
🔗 发布地址：https://creator.xiaohongshu.com/publish/publish

Douyin format:

📋 抖音发布内容：

作品描述：{title}
简介：{body_truncated_1000}
标签：#tag1 #tag2 #tag3

📎 视频：{video asset path}
📎 封面：{cover asset path or "使用自动封面"}
🔗 发布地址：https://creator.douyin.com/creator-micro/content/upload

WeChat Video format:

📋 视频号发布内容：

创作描述：
{title}
{body_truncated_900}

#tag1 #tag2 #tag3

📎 视频：{video asset path}
🔗 发布地址：https://channels.weixin.qq.com/platform/post/create

WeChat MP (公众号) format:

📋 公众号发布内容：

标题：{title}
正文：
{body_with_formatting}

📎 封面：{cover asset path or "需要手动上传"}
🔗 发布地址：https://mp.weixin.qq.com/

After user confirms they published, update status:

{ "action": "confirm_published", "content_id": "content-xxx", "publish_url": "<optional platform url>" }

---

Browser Automation Steps (OpenClaw Only)

Common: Navigate and check login

{"action": "navigate", "targetUrl": "<platform create URL>"}

Wait 5 seconds. Take snapshot.

• IF URL matches login detection pattern → tell user: "平台登录已过期，请手动登录后重试。"
• ELSE → continue to platform-specific steps.

---

Steps (Douyin)

URL: https://creator.douyin.com/creator-micro/content/upload After video upload, auto-redirects to /content/publish

D1: Upload video

Find file input and upload:

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const input = document.querySelector('input[type=file][accept*=video], input[type=file][accept*=mp4]'); if (input) { input.style.display = 'block'; return 'found'; } return 'not_found'; })()"
  }
}

Get video path from project assets, then:

{
  "action": "act",
  "request": {
    "kind": "setInputFiles",
    "selector": "input[type=file][accept*=video], input[type=file][accept*=mp4]",
    "files": ["<video_path>"]
  }
}

Wait for upload (poll every 5s, up to 120s). Check if URL changes to /content/publish.

D2: Fill form

Title (作品描述, max 30 chars):

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const input = document.querySelector('input[placeholder*=标题], input[placeholder*=作品]'); if (input) { const nativeSet = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value').set; nativeSet.call(input, '<title_30>'); input.dispatchEvent(new Event('input', {bubbles: true})); return 'filled'; } return 'not_found'; })()"
  }
}

Description (简介, max 1000 chars) — contenteditable div:

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const editors = document.querySelectorAll('[contenteditable=true]'); for (const ed of editors) { if (ed.offsetHeight > 50) { ed.focus(); ed.textContent = '<description_1000>'; ed.dispatchEvent(new Event('input', {bubbles: true})); return 'filled'; } } return 'not_found'; })()"
  }
}

Tags — append to description or find tag input:

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const editors = document.querySelectorAll('[contenteditable=true]'); for (const ed of editors) { if (ed.textContent.includes('添加话题') || ed.closest('[class*=topic]')) { ed.focus(); const tags = ['tag1','tag2','tag3']; tags.slice(0,5).forEach(t => { ed.textContent += ' #' + t; }); ed.dispatchEvent(new Event('input', {bubbles: true})); return 'added'; } } return 'not_found'; })()"
  }
}

D3: Set cover (if cover asset exists)

Find cover file input:

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const inputs = document.querySelectorAll('input[type=file][accept*=image]'); for (const inp of inputs) { const parent = inp.closest('[class*=cover], [class*=poster]'); if (parent) { inp.style.display = 'block'; return 'found'; } } return 'not_found'; })()"
  }
}

Then setInputFiles with cover path. Wait 3s.

D4: Publish

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const btns = document.querySelectorAll('button'); for (const b of btns) { if (b.textContent.trim() === '发布' && !b.disabled) { b.click(); return 'clicked'; } } return 'not_found'; })()"
  }
}

Wait 5s. Verify success (redirect to content management or success message).

---

Steps (Xiaohongshu)

URL: https://creator.xiaohongshu.com/publish/publish Tabs: 上传视频 / 上传图文 / 写长文 (tabs NOT in accessibility tree, use JS click)

X1: Switch tab if needed

IF no video (image note) → switch to "上传图文":

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const tabs = document.querySelectorAll('[class*=tab] span, [role=tab]'); for (const t of tabs) { if (t.textContent.includes('上传图文')) { t.click(); return 'switched'; } } return 'not_found'; })()"
  }
}

X2: Upload media

Video or images via input[type=file] + setInputFiles. Same pattern as Douyin.

X3: Fill form

Title (max 20 chars):

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const input = document.querySelector('input[placeholder*=标题], input[class*=title]'); if (input) { const nativeSet = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value').set; nativeSet.call(input, '<title_20>'); input.dispatchEvent(new Event('input', {bubbles: true})); return 'filled'; } return 'not_found'; })()"
  }
}

Body (max 1000 chars) — contenteditable:

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const editors = document.querySelectorAll('[contenteditable=true], textarea[placeholder*=正文]'); for (const ed of editors) { if (ed.offsetHeight > 80 || ed.tagName === 'TEXTAREA') { ed.focus(); if (ed.tagName === 'TEXTAREA') { ed.value = '<body_1000>'; } else { ed.textContent = '<body_1000>'; } ed.dispatchEvent(new Event('input', {bubbles: true})); return 'filled'; } } return 'not_found'; })()"
  }
}

Tags — append #tag1 #tag2 to body, or find tag input and Enter each tag.

X4: Publish

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const btns = document.querySelectorAll('button'); for (const b of btns) { const t = b.textContent.trim(); if ((t === '发布' || t === '发布笔记') && !b.disabled) { b.click(); return 'clicked'; } } return 'not_found'; })()"
  }
}

---

Steps (WeChat Video)

URL: https://channels.weixin.qq.com/platform/post/create

W1: Upload video — same input[type=file] + setInputFiles pattern.

W2: Fill 创作描述 — combine title + body + tags into one field (no separate title):

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const editors = document.querySelectorAll('[contenteditable=true], textarea'); for (const ed of editors) { if (ed.offsetHeight > 50) { ed.focus(); const text = '<title>\\n<body_900>\\n\\n#tag1 #tag2 #tag3'; if (ed.tagName === 'TEXTAREA') { ed.value = text; } else { ed.textContent = text; } ed.dispatchEvent(new Event('input', {bubbles: true})); return 'filled'; } } return 'not_found'; })()"
  }
}

W3: Set cover — find cover file input if custom cover exists.

W4: Toggle 原创声明 (if original content):

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const switches = document.querySelectorAll('[role=switch], input[type=checkbox]'); for (const s of switches) { if (s.closest('label')?.textContent?.includes('原创')) { if (!s.checked) { s.click(); return 'toggled'; } return 'already_on'; } } return 'not_found'; })()"
  }
}

W5: Publish — button text is "发表":

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const btns = document.querySelectorAll('button'); for (const b of btns) { if ((b.textContent.trim() === '发表' || b.textContent.trim() === '发布') && !b.disabled) { b.click(); return 'clicked'; } } return 'not_found'; })()"
  }
}

---

Steps (WeChat MP 公众号)

URL: https://mp.weixin.qq.com/cgi-bin/appmsg?t=media/appmsg_edit&action=edit&type=77 This is the "图文消息" editor. Navigate via: 内容与互动 → 图文消息 → 写新图文

M1: Navigate to new article editor

{"action": "navigate", "targetUrl": "https://mp.weixin.qq.com/cgi-bin/appmsg?t=media/appmsg_edit&action=edit&type=77"}

Wait 5s. If redirected to login → tell user to log in manually.

M2: Fill title

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const input = document.querySelector('#title, input[placeholder*=标题], input[name=title]'); if (input) { const nativeSet = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value').set; nativeSet.call(input, '<title_64>'); input.dispatchEvent(new Event('input', {bubbles: true})); return 'filled'; } return 'not_found'; })()"
  }
}

M3: Fill body

The MP editor uses a rich text iframe or contenteditable. Fill with formatted content:

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const editor = document.querySelector('#edui1_body, [contenteditable=true][class*=rich], .edui-body-container'); if (editor) { editor.innerHTML = '<formatted_html_body>'; editor.dispatchEvent(new Event('input', {bubbles: true})); return 'filled'; } const iframe = document.querySelector('#ueditor_0, iframe[class*=edui]'); if (iframe) { const doc = iframe.contentDocument || iframe.contentWindow.document; const body = doc.querySelector('body[contenteditable=true], body'); if (body) { body.innerHTML = '<formatted_html_body>'; return 'filled_iframe'; } } return 'not_found'; })()"
  }
}

Note: MP body supports HTML formatting. Convert markdown to HTML:

• ## heading → <h2>heading</h2>
• **bold** → <strong>bold</strong>
• Paragraphs → <p>text</p>
• Blank lines → <p><br></p>

M4: Set cover image

Find the cover upload area (usually at the top of the editor):

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const inputs = document.querySelectorAll('input[type=file][accept*=image]'); for (const inp of inputs) { inp.style.display = 'block'; return 'found'; } return 'not_found'; })()"
  }
}

Then setInputFiles with cover path.

M5: Preview and publish

MP has a two-step flow: "预览" then "群发" or "发布".

For draft save (safer):

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const btns = document.querySelectorAll('button, a'); for (const b of btns) { const t = b.textContent.trim(); if (t === '保存' || t === '保存草稿') { b.click(); return 'saved_draft'; } } return 'not_found'; })()"
  }
}

For direct publish:

{
  "action": "act",
  "request": {
    "kind": "evaluate",
    "fn": "(() => { const btns = document.querySelectorAll('button, a'); for (const b of btns) { const t = b.textContent.trim(); if (t === '群发' || t === '发布') { b.click(); return 'clicked'; } } return 'not_found'; })()"
  }
}

---

Common: Update Status

After publish (success or failure), update content status:

{ "action": "update", "id": "content-xxx", "status": "published" }

Output:

✅ 已发布到{platform_label}

or

❌ 发布失败：{error_message}。请检查账号状态后重试。

Selector Adaptation Strategy

DOM selectors are based on page structures observed in early 2026. Pages will change.

When a selector returns not_found:

1. Use browser snapshot to see current page structure
2. Identify the correct element by role, placeholder, or nearby labels
3. Adapt the selector
4. Core pattern stays the same: evaluate → find element → fill/click

Key hints:

• Douyin: Title placeholder "填写作品标题". Publish button "发布".
• Xiaohongshu: Title placeholder contains "标题". Publish button "发布" or "发布笔记".
• WeChat Video: One main text area (创作描述). Publish button "发表".
• WeChat MP: Title input #title. Body in iframe or contenteditable. Save "保存", publish "群发".

Error Handling

Failure	Action
Login expired (redirect)	Tell user to log in manually.
Upload timeout (>120s)	Report failure, suggest retry.
CAPTCHA/verification popup	Report failure, suggest manual login.
Form validation error	Read error from DOM, report to user.
Content too long	Truncate to limit + "…", proceed.
setInputFiles fails	Try all input[type=file] on page.
Publish button not found	Use snapshot to find correct button.
Browser not available	Fall back to copy-paste mode.

Changelog

• 2026-04-01: v2 — Added clipboard-first publishing flow (Free tier default). New actions: clipboard, confirm_published. Added bilibili platform. Tier routing: Free=clipboard, Pro=browser CDP.
• 2026-03-31: v1 — Adapted from Qingmo publish-content.md v2. Added WeChat MP (公众号) support. Dual-mode: copy-paste + browser automation. Removed backend API dependency. Assets loaded from local project directory.